Chow test and Structual breaks.

[chowchow]

Hi,

Im having a problem with running a chow test. I have a sample period of the Bovespa and Oil prices from 2000-2009. I've identified a structural break around 05/2008 for both variables by using a graph. (this is approx start of the current financial crisis) So I'm looking to get a similar break point for both my variables.

The steps I've followed are:
Run and OLS regression between Bovespa and Oil
Then Run the chow test from the stability test menu.
My results are very strange as I've ran through different stages of the sample period, and its basically saying that the entire period is a structural break.

I was doing it a different way before and it seemed to give me better results, but I was told it was incorrect.
I was running an ols between Oil and bovespa, then getting the residual series, then running an ols residuals c residuals(-1), and then doing the chow test off the residuals.

Can anyone advise me the proper way to get the break point?

Many thanks,

[EViews Gareth]

It isn't very clear what you're asking here....


The Chow test can be used to test whether there is a structural break at a specific date. It isn't very good at finding a structural break, though. I.e. you have to have a specific date in mind before testing.

The Chow test is also not very good at rejecting a structural break if the true break as in fact a close date, but not the date being tested. For example say there was actually a break in May 2006. If you tested August 2006, it is likely that the Chow test would say that there was, in fact, a break in August. This is probably why if you perform a Chow test on a lot of dates that are close together, they will all say, "Yes".


If you want to find the actual date of a structural break, you should try the Quandt-Andrews breakpoint test (available from the same menu as the Chow).

[chowchow]

Basically I have identified a possible structural break in my data using graphs. My sample period was from 2000-2009, now I'm looking to split the data. I want to carry out a chow test to find the exact location of the structural break for 2 variables, so I can split the data at the same point.
I have run an OLS regression between variable one and variable two - and then do a chow test from there, but my results seem to be inaccurate as the entire sample seems to give a structural break.

I also tried another way using the residuals of the OLS regression and this seems to give better results, as not all of the sample gives strucutral breaks but was told before that this was incorrect.
The steps carried out using the residuals are the following:
OLS regression between variable one and variable two
get the resiudals of this regreesion
then run an ols (regression c regression(-1)
then do the chow test on the regressions.

Could you advice me of a way of conducting a chow test using two variables?